Daniel Yarmoluk, AI architect and founder of GraphifyMD who is also an adjunct professor at the University of St. Thomas, joins the podcast to discuss why the context window—not raw model power—has become the real bottleneck for enterprise AI. As token consumption soars and agent workforces scale, Daniel and host Michael Gale reframe the challenge as one of architecture: most context capacity is consumed by orchestration, rules, and probabilistic RAG, leaving only a sliver for the domain knowledge that drives good decisions. They explore what metrics CFOs should really be prioritizing, such as cost-per-inference and "intelligence per watt." They also examine why knowledge graphs and ontologies, not more pattern matching, are the way to reduce hallucination and drift, especially for high-stakes industries.
